Immunohistochemistry analysis model construction method, device, equipment and storage medium

Abstract

The present invention provides a method, device, equipment and storage medium for constructing an analysis model of immunohistochemistry, which relates to the field of image analysis technology. The method includes: obtaining sample data and a positive expression threshold of each marker protein in the sample; constructing a target Gaussian mixture model of the expression of the marker protein in the cell based on the sample data; inputting the sample data into the target Gaussian mixture model to obtain the target parameter of each sample data; constructing a generalized linear model with the positive expression threshold as the response variable, and training the generalized linear model with the target parameter to obtain an analysis model of immunohistochemistry. By using the analysis model of the present application, automatic analysis of immunohistochemical staining images can be achieved, eliminating the interference of subjective factors in manual analysis; the analysis results have good consistency and high accuracy.

Description

Technical Field

[0001] The present invention relates to the field of image analysis technology, and in particular to a method, device, equipment and storage medium for constructing an immunohistochemical analysis model. Background Art

[0002] Conventional immunohistochemical staining analysis is to split cells by identifying DAP I nuclear staining, that is, marking proteins with different colors and displaying them on the same image, so as to obtain the expression of different marker proteins in each cell, and then manually define the threshold of positive cells to calculate the positive rate of different samples. This analysis scheme is heavily dependent on the subjective judgment of analysts, and it is difficult to form a fixed standard. The results obtained by different analysts often have certain deviations. Summary of the invention

[0003] The purpose of the present invention is to provide a method, device, equipment and storage medium for constructing an immunohistochemical analysis model to improve the above problems. In order to achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0004] In a first aspect, the present application provides a method for constructing an immunohistochemical analysis model, comprising:

[0005] Obtain sample data and the positive expression threshold of each marker protein in the sample;

[0006] Based on the sample data, a target Gaussian mixture model of the expression amount of the intracellular marker protein is constructed; the sample data is input into the target Gaussian mixture model to obtain a target parameter for each sample data;

[0007] A generalized linear model is constructed with the positive expression threshold as the response variable, and the generalized linear model is trained using the target parameter to obtain an immunohistochemical analysis model.

[0008] In a second aspect, the present application also provides an immunohistochemical analysis model construction device, the device comprising:

[0009] A data acquisition module, used to obtain sample data and the positive expression threshold of each marker protein in the sample;

[0010] A feature extraction module is used to construct a target Gaussian mixture model of the expression amount of the intracellular marker protein based on the sample data; the sample data is input into the target Gaussian mixture model to obtain the target parameter of each sample data;

[0011] The model building module is used to build a generalized linear model with the positive expression threshold as the response variable, and train the generalized linear model using the target parameter to obtain an immunohistochemical analysis model.

[0012] In a third aspect, the present application also provides an immunohistochemical analysis model construction device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned immunohistochemical analysis model construction method when executing the computer program.

[0013] In a fourth aspect, the present application further provides a readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned immunohistochemical analysis model construction method are implemented.

[0014] The beneficial effects of the present invention are:

[0015] The present invention constructs a Gaussian mixture model based on existing sample data to obtain target parameters, and then trains a generalized linear model in combination with the sample data, the positive expression threshold of the sample and the target parameters to construct an immunohistochemical analysis model.

[0016] By using the analysis model of the present application, automatic analysis of immunohistochemical staining images can be achieved, eliminating the interference of subjective factors in manual analysis; the analysis results of samples with similar marker protein expression distributions have good consistency and high accuracy.

[0017] Other features and advantages of the present invention will be set forth in the following description, and in part will become apparent from the description, or may be understood by practicing embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0019] Figure 1 This is a flow chart of the method for constructing an immunohistochemical analysis model in the embodiments of the present application;

[0020] Figure 2 Schematic diagram of classification of negative cells and positive cells in the embodiment of the present application;

[0021] Figure 3 This is a schematic diagram of the structure of the immunohistochemical analysis model construction device in the embodiment of the present application;

[0022] Figure 4 This is a schematic diagram of the structure of the immunohistochemical analysis model construction device in the embodiments of the present application.

[0023] Markings in the figure: 1. Data acquisition module; 2. Feature extraction module; 3. Model building module; 800. Immunohistochemistry analysis model building device; 801. Processor; 802. Memory; 803. Multimedia component; 804. I / O interface; 805. Communication component. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0025] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0026] Multiple immunohistochemical staining can display the expression of proteins in different colors on the same image, which helps to judge the distribution of multiple marker proteins in tissues in clinical applications. By analyzing each layer of the multiple immunohistochemical staining image, the positive rate and density information of various marker proteins expressed in the sample can be obtained.

[0027] Conventional multiple immunohistochemistry (mI HC) staining image analysis uses professional software to obtain the expression levels of different marker proteins in each cell type, and then manually defines the positive cell threshold to calculate the positive rate of different samples. This analysis method is heavily dependent on the subjective judgment of analysts and requires manual analysis, recording, and statistics of each sample and each layer. The operation is cumbersome and prone to errors.

[0028] Example 1

[0029] like Figure 1 As shown, this embodiment provides a method for constructing an immunohistochemical analysis model, which is suitable for analyzing multiple immunohistochemical staining images. The method includes step S100, step S200 and step S300.

[0030] Step S100, obtaining sample data and a positive expression threshold of each marker protein in the sample; the marker protein is a fluorescently labeled protein;

[0031] The sample data obtained in this step is data directly output by professional software after the immunohistochemical staining slice sample is subjected to cell identification through professional software. The professional software may specifically be QuPath, HALO (digital pathology image analysis platform), etc.

[0032] In this embodiment, the sample data includes the expression level of each protein in the nucleus, cytoplasm and the whole cell of each cell in the sample. According to the acquired sample data, a corresponding positive expression threshold is manually defined for each marker protein, and cells with an expression level higher than or equal to the positive expression threshold are identified as positive cells, and cells with an expression level lower than the positive expression threshold are identified as negative cells. Cells can also be binarized so that positive cells and negative cells are displayed in different colors, such as Figure 2 shown.

[0033] Step S200: constructing a target Gaussian mixture model of the expression of intracellular marker proteins based on the sample data; inputting the sample data into the target Gaussian mixture model to obtain target parameters for each sample data;

[0034] In this embodiment, before constructing the target Gaussian mixture model, the sample data is also preprocessed, including:

[0035] Screening the sample data, eliminating data corresponding to samples whose cell numbers are less than a preset number threshold; and calculating the occurrence rate of each marker protein in the sample, eliminating data corresponding to samples where the marker protein occurrence rate is less than a preset ratio threshold, to obtain screening data;

[0036] The data were eliminated based on the above two conditions in order to prevent samples with too few cells and samples containing rare proteins from affecting model building. As an example, the quantity threshold can be set to 5000, that is, samples with less than 5000 cells will be eliminated; the ratio threshold can be set to 1% of the total amount of marker proteins in all samples, that is, if the occurrence rate of a certain marker protein in all samples is less than 1%, it can be considered as a rare protein, and samples containing the rare protein will be eliminated.

[0037] After obtaining the screened data, a portion of the data is randomly selected to construct the target Gaussian mixture model;

[0038] In this embodiment, the random selection method is: desensitize the screening data, shuffle the sample order and renumber them, select the first 80% of the samples as the training set, and use them to build the target Gaussian mixture model. The remaining 20% ​​of the samples after random selection can be used as a test set to test the constructed analysis model in subsequent steps.

[0039] To construct a target Gaussian mixture model of intracellular marker protein expression, it is first necessary to construct an initial Gaussian mixture model based on the preprocessed sample data. The method for constructing the initial Gaussian mixture model is as follows:

[0040] Classifying cells according to the expression amount of the marker protein to obtain positive cells and negative cells;

[0041] The sample data output by general professional software includes the expression of the marker protein in the cell nucleus, cytoplasm and the whole cell at the same time, that is, one marker protein in one cell corresponds to three expression data, but only one of the data can be used as the expression level (expression amount) of the marker protein;

[0042] Therefore, it is necessary to first obtain the type information of the marker protein, and determine the expression pattern of the marker protein based on the type information; the expression pattern includes nuclear expression, cytoplasmic expression and whole cell expression;

[0043] Then, according to the expression pattern of the marker protein, data matching the expression pattern is screened out from the sample data to obtain the expression amount of the marker protein in the cell.

[0044] If marker protein A is mainly expressed in the cell nucleus, the expression data of the marker protein in the cell nucleus should be selected as the marker protein expression in the cell, and the expression data of A in the cytoplasm should not be selected as the marker protein expression in the cell.

[0045] After determining the expression level of the marker protein in the cell, it can be considered that the expression level of the marker protein in the positive cells and the negative cells obeys the Gaussian distribution respectively, and the formula is:

[0046]

[0047] Among them, G(x) is the probability density of the expression of positive cell or negative cell marker protein, σ is the standard deviation of the Gaussian function, and μ is the expectation of the Gaussian function.

[0048] Constructing a first independent Gaussian model according to the marker protein expression level of the positive cells;

[0049] Constructing a second independent Gaussian model according to the expression level of the marker protein of the negative cells;

[0050] The first independent Gaussian model and the second independent Gaussian model are superimposed as sub-models to obtain the initial Gaussian mixture model, which is:

[0051]

[0052] Among them, G(x;θ) is the probability density of the marker protein expression in the whole sample, α k represents the probability of the kth sub-model appearing in the population, K represents the total number of sub-models, and in this application, K=2. represents the kth sub-model, θ represents the parameter set of each sub-model, θ k Represents the parameters of the kth sub-model.

[0053] Since both positive cells and negative cells exist in the same sample, the first independent Gaussian model and the second independent Gaussian model need to be superimposed so that the resulting Gaussian mixture model can display the probability density of the marker protein expression in the sample.

[0054] In order to obtain the target Gaussian mixture model, the parameters θ of the initial Gaussian mixture model need to be k To solve, in this embodiment, the maximum expectation algorithm is used to train the initial Gaussian mixture model to solve θ k .

[0055] The Expectation Maximization (EM) algorithm is known as one of the top ten machine learning algorithms. It is a method for maximum likelihood estimation of probability model parameters from incomplete data or data sets with missing data (there are hidden variables).

[0056] The EM algorithm includes E-step and M-step, where E-step represents the expected step of each iteration and M-step represents the maximum step.

[0057] The E-step needs to first assume the parameters of a sub-model, namely the mean and standard deviation, and then calculate whether each cell is more likely to belong to the first or second sub-model, specifically:

[0058] Initialize θ k , that is, inputting the initial values ​​of μ and σ; the method for determining the initial values ​​of μ and σ in this embodiment includes:

[0059] Binning the sample data according to the value to obtain a binned data set;

[0060] Arrange the binned data sets by size, and remove the largest and smallest binned data sets;

[0061] The sample data is randomly selected from the remaining binned data sets, and the expected initial value of μ and the initial value of the standard deviation σ of the initial Gaussian mixture model are determined according to the randomly selected sample data.

[0062] Let p ik Represents the probability of the i-th cell in the sample in the k-th sub-model;

[0063] Calculate p based on the initial values ​​of μ and σ ik :

[0064]

[0065] M-Step re-estimates the Gaussian distribution parameters of the two sub-models after the initial classification of cells based on the above calculations:

[0066] Further calculate the expected μ of the kth sub-model k :

[0067]

[0068] Further calculate the standard deviation σ of the kth sub-model k :

[0069]

[0070] The obtained μ k and σ k Substitute the initial values ​​of μ and σ and iterate until |θ j -θ j+1 |<ε, terminate the iteration and get the final θ k , where j represents the number of iterations and ε is the preset minimum value for judging convergence.

[0071] Get the final θ k Then the target Gaussian mixture model is obtained, and each sample can obtain two corresponding groups (negative cells and positive cells) of target parameters σ and μ, which are used as input parameters for subsequent analysis model training.

[0072] Step S300: construct a generalized linear model using the positive expression threshold as a response variable, and train the generalized linear model using the target parameter to obtain an immunohistochemical analysis model.

[0073] The positive expression threshold manually defined in step S100 is used as the response variable to construct a generalized linear model, and the two groups of target parameters σ and μ obtained in step S200 are used as independent variables to train the generalized linear model to obtain an immunohistochemical analysis model.

[0074] As an optional implementation, in order to verify the immunohistochemical analysis model constructed in the present application, the present application further includes step S400:

[0075] Based on the positive expression threshold, a standard positive rate of the sample is obtained;

[0076] Inputting the preset test data into the analysis model to obtain the predicted positive rate of the test data;

[0077] The deviation between the predicted positive rate and the standard positive rate is calculated to verify the analysis model.

[0078] The test data, namely, two groups of α and μ of the test set calculated based on the test set, may be sample data corresponding to the remaining 20% ​​of the samples after the random selection in the above step S200.

[0079] In this embodiment, if the deviation between the predicted positive rate and the standard positive rate does not exceed ±10%, it is considered that the analysis model meets the requirements.

[0080] This example uses real data of multi-color immunohistochemical staining images of a company. After preprocessing the sample data, 875 samples were randomly selected. Each sample contains expression data of 2 to 7 marker proteins, a total of 4537 groups of marker protein expression data, involving 33 panels (panel) and 65 protein markers.

[0081] 80% of the samples were randomly selected as the training set, which contained 700 samples in total, 3683 groups of marker protein expression data, involving 26 panels and 51 protein markers; an analysis model was constructed based on the data of the validation set through the steps of the above method;

[0082] The remaining samples were used as the test set, which contained 175 samples in total and 854 sets of marker protein expression data, involving 10 panels and 34 protein markers. The test set was input into the constructed analysis model, and the deviation between the output predicted positive rate and the standard positive rate of manual analysis was 4.8%, with high analysis accuracy.

[0083] Example 2

[0084] like Figure 3 As shown, this embodiment provides an immunohistochemical analysis model construction device, the device comprising:

[0085] Data acquisition module 1, used to obtain sample data and the positive expression threshold of each marker protein in the sample;

[0086] Feature extraction module 2, used to construct a target Gaussian mixture model of the expression of intracellular marker proteins based on the sample data; input the sample data into the target Gaussian mixture model to obtain target parameters for each sample data;

[0087] The model building module 3 is used to build a generalized linear model with the positive expression threshold as the response variable, and train the generalized linear model using the target parameter to obtain an immunohistochemical analysis model.

[0088] In a specific embodiment disclosed in the present application, the analysis model building device further includes:

[0089] The verification module is used to obtain the standard positive rate of the sample based on the positive expression threshold; input the preset test data into the analysis model to obtain the predicted positive rate of the test data; calculate the deviation between the predicted positive rate and the standard positive rate, and verify the analysis model.

[0090] In a specific embodiment disclosed in the present application, the feature extraction module includes:

[0091] A data processing unit is used to screen the sample data, remove the data corresponding to the samples whose cell numbers are less than a preset number threshold; and calculate the occurrence rate of each marker protein in the sample, remove the data corresponding to the samples where the marker protein occurrence rate is less than a preset ratio threshold, and obtain screening data;

[0092] The Gaussian model building unit is used to build a target Gaussian mixture model of the expression amount of the intracellular marker protein based on the screening data.

[0093] Example 3

[0094] Corresponding to the above method embodiment, this embodiment also provides an immunohistochemical analysis model construction device. The immunohistochemical analysis model construction device described below and the immunohistochemical analysis model construction method described above can refer to each other.

[0095] Figure 4 FIG. 8 is a block diagram of an immunohistochemical analysis model building device 800 according to an exemplary embodiment. Figure 3As shown, the analysis model building device 800 includes a processor 801 and a memory 802. The analysis model building device 800 may also include one or more of a multimedia component 803, an input / output (I / O) interface 804, and a communication component 805. Among them, the processor 801 is used to control the overall operation of the analysis model building device 800 to complete all or part of the steps in the above-mentioned analysis model building method. The memory 802 is used to store various types of data to support the operation of the analysis model building device 800, which may include, for example, commands for any application or method operating on the analysis model building device 800, as well as application-related data, such as contact data, messages sent and received, pictures, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination of them, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, disk or optical disk.

[0096] The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone, and the microphone is used to receive external audio signals.

[0097] The received audio signal can be further stored in the memory 802 or sent through the communication component 805. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules can be keyboards, mice, buttons, etc. These buttons can be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the analysis model building device 800 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 805 can include: Wi-Fi module, Bluetooth module, NFC module.

[0098] In an exemplary embodiment, the device 800 for mutual signing and verification of digital files can be implemented by one or more application-specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned analysis model construction method.

[0099] In another exemplary embodiment, a computer-readable storage medium including program commands is also provided, and when the program commands are executed by a processor, the steps of the above-mentioned analysis model construction method are implemented. For example, the computer-readable storage medium can be the above-mentioned memory 802 including program commands, and the above-mentioned program commands can be executed by the processor 801 of the analysis model construction device 800 to complete the above-mentioned analysis model construction method.

[0100] Example 4

[0101] Corresponding to the above analysis model construction method embodiment, this embodiment further provides a readable storage medium, and the readable storage medium described below and the analysis model construction method described above can refer to each other.

[0102] A readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the analysis model construction method of the above-mentioned analysis model construction method embodiment.

[0103] The readable storage medium may specifically be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or other readable storage medium that can store program codes.

[0104] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0105] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for constructing an immunohistochemical analysis model, characterized in that: include: Obtain sample data and the positive expression threshold of each marker protein in the sample; Based on the sample data, a target Gaussian mixture model of the expression of intracellular marker proteins is constructed, including: Based on the sample data, an initial Gaussian mixture model is constructed, and the initial Gaussian mixture model is trained by a maximum expectation algorithm to obtain the target Gaussian mixture model; The initial Gaussian mixture model is constructed based on the sample data, comprising: Classifying cells according to the expression amount of the intracellular marker protein to obtain positive cells and negative cells; Constructing a first independent Gaussian model according to the marker protein expression level of the positive cells; Constructing a second independent Gaussian model according to the expression level of the marker protein of the negative cells; Superimposing the first independent Gaussian model and the second independent Gaussian model to obtain the initial Gaussian mixture model; Inputting the sample data into the target Gaussian mixture model to obtain target parameters for each sample data; A generalized linear model is constructed with the positive expression threshold as the response variable, and the generalized linear model is trained using the target parameter to obtain an immunohistochemical analysis model.

2. The method for constructing an immunohistochemical analysis model according to claim 1, characterized in that: The analysis model building method further includes: Based on the positive expression threshold, a standard positive rate of the sample is obtained; Inputting the preset test data into the analysis model to obtain the predicted positive rate of the test data; The deviation between the predicted positive rate and the standard positive rate is calculated to verify the analysis model.

3. The method for constructing an immunohistochemical analysis model according to claim 1, characterized in that: The step of constructing a target Gaussian mixture model of the expression of intracellular marker proteins based on the sample data comprises: Screening the sample data, eliminating data corresponding to samples with cell numbers less than a preset threshold value; and calculating the occurrence rate of marker proteins in the samples, eliminating data corresponding to samples with marker proteins with an occurrence rate less than a preset ratio threshold value, to obtain screening data; Based on the screening data, a target Gaussian mixture model of the marker protein expression level in cells is constructed.

4. An immunohistochemical analysis model construction device, characterized in that: The device comprises: A data acquisition module, used to obtain sample data and the positive expression threshold of each marker protein in the sample; A feature extraction module is used to construct a target Gaussian mixture model of the expression amount of intracellular marker proteins based on the sample data, including: Based on the sample data, an initial Gaussian mixture model is constructed, and the initial Gaussian mixture model is trained by a maximum expectation algorithm to obtain the target Gaussian mixture model; The initial Gaussian mixture model is constructed based on the sample data, comprising: Classifying cells according to the expression amount of the intracellular marker protein to obtain positive cells and negative cells; Constructing a first independent Gaussian model according to the marker protein expression level of the positive cells; Constructing a second independent Gaussian model according to the expression level of the marker protein of the negative cells; Superimposing the first independent Gaussian model and the second independent Gaussian model to obtain the initial Gaussian mixture model; Inputting the sample data into the target Gaussian mixture model to obtain target parameters for each sample data; The model building module is used to build a generalized linear model with the positive expression threshold as the response variable, and train the generalized linear model using the target parameter to obtain an immunohistochemical analysis model.

5. The immunohistochemical analysis model construction device according to claim 4, characterized in that The device also includes: Verification module: used to obtain the standard positive rate of the sample based on the positive expression threshold; input the preset test data into the analysis model to obtain the predicted positive rate of the test data; calculate the deviation between the predicted positive rate and the standard positive rate, and verify the analysis model.

6. The immunohistochemical analysis model construction device according to claim 4, characterized in that: The feature extraction module comprises: A data processing unit is used to screen the sample data, remove the data corresponding to the samples whose cell numbers are less than a preset number threshold; and calculate the occurrence rate of each marker protein in the sample, remove the data corresponding to the samples where the marker protein occurrence rate is less than a preset ratio threshold, and obtain screening data; The Gaussian model building unit is used to build a target Gaussian mixture model of the expression amount of the intracellular marker protein based on the screening data.

7. An immunohistochemical analysis model construction device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 3 are implemented.

8. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images